Our perception of the world is mediated by logical maps, meant to make sense of the river of raw information processed by the senses. When it comes to architecture, we learn to describe and think through ideas such as interior and exterior, floor, ceiling and walls, structure and skin.
While this is mostly valuable to operate, communicate and engage with the industry at all levels, it is also a restriction on the creation of new visions. As designer you would want to shake some of these biases off, and explore those darker territories, where things can be hybrid and ambiguous, as in dreams and hallucinations, and possibly come back with something new.
At the moment of writing, August 2022, artificial intelligence is giving a chance for taking on the challenge. Many of the images produced by tools as Midjourney AI retain a level of uncertainty, descending from the fact that the network is making operations with pixels and text, while not having spatial, structural or material understanding.
Even though such productions can be very convincing at a first glance, they often include wrong perspectives, an inconsistent definition of depth, features that merge into others while breaking the standard rules of drawing and imaging, structural incoherencies. The first reaction is to perceive them as errors, ignore or correct them.
With this series of images I tried to do something different, to explore a visual language and aesthetics that accentuates the visual mistakes and illusions, rather than correct them, by orienting the process of co-creation with an AI, selection and iteration, to evolve the images with the less straightforward reading.
This opens the possibility of pursuing a dreamlike, surreal quality, that would otherwise be difficult to achieve with 3d modelling and rendering, where every step forces the designer to apply a logical structure and high level of coherence.
All images produced by Filippo Nassetti using Midjourney AI.
Midjourney AI is a form of artificial intelligence and application that produces images in response to a text input. The user prompts a text description, and the network answers with a few possible imaginations, that can be progressively refined.
Back to Top